Hi,

I have a question about an analysis of RNA seq data via DESeq2.
The RNA seq design is the following:

Sample                 experiment              siRNA                   hormonal treatment             

1                             M0                               sineg                     control

2                             M1                               sineg                     control

3                             M2                               sineg                     control

4                             M0                               sineg                     hormone

5                             M1                               sineg                     hormone

6                             M2                               sineg                     hormone

7                             M0                               siX                          control

8                             M1                               siX                          control

9                             M2                               siX                          control

10                           M0                               siX                          hormone

11                           M1                               siX                          hormone

12                           M2                               siX                          hormone

So there are two conditions (siRNA and hormonal treatment) and within each condition there are a control sample (sineg or control) and a treated sample (siX or hormone). And we realized 3 experiments (M0,M1,M2).

I would like to identify genes that respond differently to hormonal treatment in condition siX compared to the condition sineg. The design I used is the following: ~ hormonal treatment + si + hormonal treatment :si. Is it OK?

Finally I would like to take into account the experiment. In fact my experiments are paired. For example, a treatment for the experiment M0 must be compared with the control of the same experiment M0. So my question is: how do I set up the design for that?

Thanks in advance.

Severine

hi Severine,

The design to use is then:

~ experiment + hormonaltreatment + siRNA + hormonaltreatment:siRNA

This will control for the experiment effect, so taking into account the pairing of samples.

As the interaction of interest is between two factors both with only two levels, I'd recommend the following:

dds = DESeq(dds, modelMatrixType="standard")
results(dds)

(The standard model matrix vs expanded doesn't make a big difference, but makes the analysis simpler in this case, reducing the interaction to a single term.)

You might also consider renaming the variables to reduce your keystrokes:

siRNA with levels neg, X

hormone with levels control, treated

And don't forget to set the base level for all factors (see vignette).